| home | help | back | first | fref | pref | prev | next | nref | lref | last | post |
Date: Thu, 7 Aug 2003 18:49:42 -0400 Message-Id: <200308072249.h77MnglR022333@pothole.mit.edu> From: "andrew m. boardman" <amb@MIT.EDU> To: bugs@MIT.EDU >So, first of all, a big "me, too" to Angie's bug about the installer >barfing on signal 11 (I think it was) when you attempt to use the >"Advanced" IP address page without guessing first. I'll look at this. >I installed a machine last night using the CD image from the bootkit >locker, and the athena cell (default) installer. The install went >great, until it rebooted, and left me with the infamous > >LI- > >prompt. For the record, I installed onto hdb5 (with hdb1 as a boot >partition), and my bootloader lives on the MBR of hda (and Windows lives >on that drive too). I don't have any weird disk geometry (I think), and >I've never had to pass custom values to LILO before - it "just worked". >Additionally, I thought the installer was supposed to use Grub? Perhaps >it didn't get a long with the LILO that was in my MBR from a plain >vanilla install of RedHat 7.2. There are a couple of problems here. First off, yes, the installer uses grub, and you probably have a perfectly functional grub install over on the MBR of hdb where it's no use whatsoever. The old LILO install is trying to load the kernel you used to have and failing. The installer assumes (and always has) that the boot partition is on your boot disk. There's no good way to work around this, but here are some not-so-good ones: - for a new install, select the "give me a shell after the install is done" option and, when given that shell, run "/sbin/grub-install /dev/<disk>", where the latter is where you want the MBR written. - for an existing but losing install that you don't want to redo, you can run an install to the point of the "Custom y/n?" query, switch vty's with Alt-F2, and do something like: # mkdir /sysimage # mount /dev/<root> /sysimage # mount /dev/<boot> /sysimage/boot # chroot /sysimage /bin/sh # /sbin/grub-install /dev/<disk> # exit # umount /sysimage/boot # umount /sysimage ...then reset the machine. The latter technique is generally useful for dealing with machines that you can't or don't want to boot. >Asking on -c consult indicates that this is a fairly common problem when >installing on hdb instead of hda. If so, can the installer perhaps >indicate that the world might break when the user selects hdb as their >drive. And/or give the option to make a bootdisk? (even a low-rent >bootdisk that is just the kernel dd'd to a floppy with rdev set >appropriately would be useful). Both a warning and the addition of an option to install the MBR on another disk would be trivial and should get added shortly; an option to create a boot disk is somewhat less so. Do you think this would still be a useful thing to be able to do if you had an installer that installed an MBR correctly in the first place? thanks for reporting this, andrew
| home | help | back | first | fref | pref | prev | next | nref | lref | last | post |